[OSM-dev] Tuning render_list: beware of hyperthreading and RAID cards latency...
cquest at openstreetmap.fr
Thu Jul 17 22:29:32 UTC 2014
Tested again with HT disabled in the bios.
Paul, you're right, it is not related to HT itself, I get more or less the
It is more related to IO bottleneck impacting more the kernel scheduling.
It may not be the same at other zoom levels where the CPU is the bottleneck
and not the IOs.
I'll switch to another topic... optimising render_list to better take
advantage of data present in the cache.
2014-07-17 17:03 GMT+02:00 Paul Norman <penorman at mac.com>:
> On 2014-07-17 6:39 AM, Christian Quest wrote:
> I'm currently doing some rendering benchmark to tune a new tile server
> we're about to put online.
> A strange thing I've found last night: render_list with 16 thread takes
> more time than with 8 .
> The server has 2 quad-core Xeon (X5570), this means 8 physical cores. It
> looks like hyperthreading adds additionnal context switches or something
> With "-z 9 -Z 9 -n 16 -l 32" it takes 3016s, and with "-z 9 -Z 9 -n 8 -l
> 12" 2647s, 10% less.
> I'll disable HT on my next run in the server bios and see if it is
> Is there a chance you're running into IO bottlenecks?
> Another thing I found... do not put your SSD behind a RAID card.
> The RAID card (Dell Perc 6i) add too much latency.
> On the same render_list, the iowait threads are increasing a lot and iops
> decreases a lot.
> The Perc 6 is an older lower-end RAID card using a LSI 10xx chip. It
> doesn't surprise me it doesn't work ideally with SSDs. I'm also not sure if
> it supports all the SATA commands a SSD uses, which can be a serious issue
> for consistency, the bane of SSDs.
> A more modern card would probably perform better, but if you're not
> RAIDing SSDs, you should really be using any card you have in HBA mode,
> unless you have a battery-backed cache and can therefor tune write barriers
> and related settings better.
Christian Quest - OpenStreetMap France
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the dev